Abstract: To acquire dynamic characteristics in brain information processing,a G-P algorithm based on the idea of restructuring of phase space is adopted.The rule of selecting three important parameters(non-graduation area,embedding dimension and delay) according to G-P algorithm is discussed.EEG signals under different brain condition are recorded and the correlation dimension is calculated.Experiments showed that the correlation dimension of EEG signal can reflect the active conditions of neuronal groups during brain information processing and can be used as the non-linear parameter of brain information processing.